III:Small: Commugrate -- A Community-based Data Integration System

该项目的目标是构建 Commugrate，这是一个社区驱动的数据集成系统，该系统利用人类社区与数据源交互中获得的信息。Commugrate 解决了科学、工程和工业中使用的信息源数量增加所带来的关键挑战，以及对大规模数据集成解决方案的需求，以实现对这些源的有效访问。这些挑战包括模式匹配和映射、记录链接和数据修复。更具体地说，Commugrate (i) 利用来自不同类型人类社区的直接和间接贡献，重点关注后者的贡献，(ii) 使用以前未使用过的新证据（例如使用情况和行为数据）解决关键数据集成问题，(iii) 采用一种新的模式匹配技术，定义了自己的一个新类别，即基于使用的模式匹配，(iv) 引入了第一个基于基于使用的记录链接的流派技术实体的行为，以及（v）提供自适应反馈系统，通过充分利用用户反馈来提高数据质量。通信对社会多个领域产生广泛影响，因为数据集成是迄今为止最重要的问题，同时也是科学、工程和工业许多领域中最棘手的问题。此外，利用用户与数据源的交互，尤其是间接交互，可以提供多种好处，并有助于解决许多棘手的数据集成任务，这些任务在没有人工干预的情况下无法完成。博士生将在该项目中进行研究。本研究的出版物、技术报告、软件和实验数据将通过该项目网站 http://www.purdue.edu/cybercenter/commugrate 传播。

The goal of this project is to build Commugrate, a community-driven data integration system that capitalizes on the information gainedfrom the interactions of communities of humans with data sources.Commugrate tackles key challenges raised by the increase in thenumber of information sources used in science, engineering, andindustry, as well as the need for large-scale data integrationsolutions to enable effective access to these sources. These challengesinclude schema matching and mapping, record-linkage, and data repair.More specifically, Commugrate (i) utilizes both direct and indirectcontributions from different types of human communities with a focus onthe latter contributions, (ii) solves key data integration issues using new evidences like usage and behavior data which have not been previously used, (iii) adopts a new technique for schema matching, which defines a new class of its own, namely usage-based schema matching, (iv) introduces the first of its genre technique for record linkage based on entities' behavior, and (v) provides an adaptive feedback system to improve the quality of the data by making the best use of users feedback. Commugrate has a broad impact across multiple segments of society as data integration is by far the most important and in the same time vexing issue in many areas in sciences, engineering, and industry. Furthermore, leveraging users' interactions with data sources, especially indirect interaction, may provide several benefits and help solve many intractable data integration tasks which cannot be done without human intervention. PhD students will pursue research in this project. Publications, technical reports, software and experimental data from this research will be disseminated via the project web site at http://www.purdue.edu/cybercenter/commugrate.